A conventional network architecture has a two-layer or multi-layer structure, and according to a function and a deployment location of each layer, a network may be divided into different layers (including a core layer, a convergence layer, an edge access layer, and a peer layer), where the layers are generally connected to one another by means of dual-homing redundant connections to improve reliability and converge user traffic layer by layer. The access layer is responsible for network access of a user terminal, is used to provide various types of user interfaces, and has widely distributed nodes and a large number of interfaces; the convergence layer is responsible for converging traffic of access nodes, is used to extend a service scope of a core node, and has various interface types, a strong convergence capability, and an integrated service processing capability; and the core layer is responsible for high-speed forwarding and service interworking between regions, and has a small number of nodes.
To integrate and simplify a network structure and improve resource utilization, in an existing network architecture, during point-to-multipoint packet switching, a method of forming a network virtualization system by using two or more main switching devices at a same network layer is used, thereby reducing complexity caused by use of multiple redundant components, and reducing a cost.
In addition, currently, to adapt to a huge capacity of data exchange between devices in a cluster, multiple high-end routers are connected to a central switch matrix chassis by using a very short reach optical fiber, to form a network virtualization system, namely, a cluster router, which includes multiple routers, thereby facilitating centralized management.
Therefore, how to enable the foregoing network virtualization system to support a multicast service becomes a technical problem to be resolved currently.